Methodology
How we test
Jokes in the intro. Science in the graph.
The short version
We change one setting, run the same test three times with it and three times without, report the median and the 1% lows, and only call it a difference if it is bigger than the normal run-to-run spread.
The test laptop
| Laptop | ASUS ROG Strix G15 |
|---|---|
| GPU | NVIDIA GeForce RTX 3060 Laptop GPU (6 GB) |
| CPU | AMD Ryzen 9 5900HX |
| RAM | 32 GB |
| Storage | 1.5 TB SSD |
| OS | Windows 11. The exact build and GPU driver are listed in each video. |
Every result applies to this laptop. Different hardware, drivers, Windows builds and games can behave differently. If you try a tweak yourself, measure it on your own machine.
Keeping it fair
- One change at a time. Only the tweak being tested changes between the two states.
- Same conditions. Plugged in, the same laptop performance mode and GPU mode for every run, the same in-game settings (for game tests: 1080p, the game's own preset, V-Sync and frame limiters off).
- Reboot between states, so leftovers from the previous setting don't leak in.
- Heat soak first. The laptop warms up (about 30 minutes for game tests) before measured runs, so we measure its real, hot performance.
- No recording during measured runs. B-roll footage is captured in a separate pass so screen capture can't eat into the numbers.
- We log the boring stuff: GPU power and temperature, room temperature, Windows build and driver version.
What we measure
- Average FPS from the median of three runs.
- 1% lows: the average of the slowest 1% of frames. This is where stutter shows up, and where some tweaks help or hurt even when the average doesn't move.
- Non-game tweaks (like startup apps or visual effects) are judged on what they claim to fix: boot time, app launch time and idle CPU, RAM and disk load, using the same median and spread rules. For these, lower is better.
How a verdict is decided
First we measure how much the runs wobble on their own. Then:
- Legit ✅ the change beats the spread and is 5% or more.
- Tiny W 🤏 the change beats the spread but is under 5%.
- Lame 🥱 the change is inside the wobble.
- Backfired 💥 it got measurably worse.
- Spicy 🌶️ an extra tag for tweaks with a security or stability cost, whatever the numbers say.
Jokes never replace a number, and we never announce a result before it has been measured.
Safety first
- A System Restore point before every test, and a full disk image before anything risky like debloat scripts.
- Security-related settings (like Memory Integrity) are switched back on after testing, on camera.
- We never run random "FPS booster" downloads or tweak packs. Fake tweak tools have been caught spreading malware, and we don't want that for you either.
Sources and disagreements
Each video's description lists the articles and tests we're checking claims against. If you think we got something wrong, tell us at hello@framesorlame.com.